* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Magnetic plasmon resonance - The University of Texas at Austin
Alternating current wikipedia , lookup
Wireless power transfer wikipedia , lookup
History of electromagnetic theory wikipedia , lookup
Superconducting magnet wikipedia , lookup
Electron paramagnetic resonance wikipedia , lookup
Maxwell's equations wikipedia , lookup
Electroactive polymers wikipedia , lookup
Magnetic field wikipedia , lookup
Neutron magnetic moment wikipedia , lookup
Electromotive force wikipedia , lookup
Magnetic nanoparticles wikipedia , lookup
Hall effect wikipedia , lookup
Faraday paradox wikipedia , lookup
Electric machine wikipedia , lookup
Electrostatics wikipedia , lookup
Electromagnetism wikipedia , lookup
Magnetic monopole wikipedia , lookup
Electricity wikipedia , lookup
History of electrochemistry wikipedia , lookup
Superconductivity wikipedia , lookup
Lorentz force wikipedia , lookup
Force between magnets wikipedia , lookup
Eddy current wikipedia , lookup
Magnetic core wikipedia , lookup
Scanning SQUID microscope wikipedia , lookup
Magnetoreception wikipedia , lookup
Magnetohydrodynamics wikipedia , lookup
Magnetochemistry wikipedia , lookup
History of geomagnetism wikipedia , lookup
PHYSICAL REVIEW E 73, 036609 共2006兲 Magnetic plasmon resonance Andrey K. Sarychev Ethertronics Inc., San Diego, California 92121, USA Gennady Shvets Department of Physics, The University of Texas at Austin, Austin, Texas 78712, USA Vladimir M. Shalaev School of Electrical and Computer Engineering, Purdue University, West Lafayette, Indiana 47907, USA 共Received 20 September 2004; revised manuscript received 23 August 2005; published 15 March 2006兲 It is demonstrated that metallic horseshoe-shaped 共also referred to as u-shaped兲 nanostructures can exhibit a magnetic resonance in the optical spectral range. This magnetic plasmon resonance is distinct from the purely geometric LC resonance occurring in perfectly conducting split rings because the plasmonic nature of the metal plays the dominant role. Similarly to the electrical surface plasmon resonance, the magnetic plasmon resonance is determined primarily by the metal properties and nanostructure geometry rather than by the ratio of the wavelength and the structure’s size. Magnetic plasmon resonance occurs in nanostructures much smaller in size than the optical wavelength. Electromagnetic properties of periodically assembled horseshoe-shaped nanostructures are investigated, and the close proximity of the electrical and magnetic plasmon resonances is exploited in designing a negative index metamaterial. Close to the magnetic plasmon resonance frequency both magnetic permeability and electric permittivity can become negative, paving the way for the development of subwavelength negative index materials in the optical range. DOI: 10.1103/PhysRevE.73.036609 PACS number共s兲: 41.20.Cv, 42.70.Qs, 42.25.Bs, 71.45.Gm I. INTRODUCTION Extending the range of electromagnetic properties of naturally occurring materials motivates the development of artificial 共or meta兲 materials. For example, it has recently been demonstrated that metamaterials may exhibit such interesting properties as negative dielectric permittivity ⬍ 0 共see, for example, Refs. 关1,2兴兲, negative magnetic permeability ⬍ 0 关3兴, and even both 关4,5兴. The double-negative case of ⬍ 0 and ⬍ 0 共often referred to as the left-handed and negative refractive regime兲 is particularly interesting because of the possibility of making a “perfect” lens with subwavelength spatial resolution 关6兴. In addition to the super resolution, unusual and sometimes counter-intuitive properties of negative index materials 共NIMs兲, which are also referred to as left-handed materials 共LHMs兲, make them very promising for applications in resonators, waveguides, and other microwave and optical elements 关7–10兴. Negative refraction has been convincingly demonstrated in the microwave regime 关5,8,9,11兴. For microwave NIMs, artificial magnetic elements 共providing ⬍ 0兲 are the splitring resonators or the swiss roll structures. In the microwave part of the spectrum, metals can be considered as perfect conductors because the skin depth is much smaller than the metallic feature size. The strong magnetic response is achieved by operating in the vicinity of the LC resonance of the split ring 关3,12兴. The same technique of obtaining ⬍ 0 using split rings was recently extended to mid-IR 关12兴 by scaling down the dimensions of the split rings. In the microwave, as well as 共to a lesser degree兲 in the mid-IR part of the spectrum, metals can be approximated as perfect conductors because the skin depth is much smaller than the feature size of the structure. Therefore, the frequencies of the LC reso1539-3755/2006/73共3兲/036609共10兲/$23.00 nances are determined entirely by the split ring geometry and size but not by the electromagnetic properties of the metal. In accordance with this, the ring response is 共resonantly兲 enhanced at some particular ratio of the radiation wavelength and the structure size. Thus we refer to the LC resonances of perfectly conducting metallic structures as geometric LC 共GLC兲 resonances. The situation drastically changes in the optical part of the spectrum, where thin 共subwavelength兲 metal components behave very differently when their sizes become less than the skin depth. For example, the electrical surface plasmon resonance 共SPR兲 occurs in the optical and near-IR parts of the spectrum due to collective electron oscillations in metal structures. Many important plasmon-enhanced optical phenomena and applications of metal nanocomposites are based on the electrical SPR 共see, for example, Ref. 关13兴兲. Plasmonic nature of the electromagnetic response in metals for optical–mid-IR frequencies is the main reason why the original methodology of GLC resonances in the microwave– mid-IR spectral range is not extendable to higher frequencies. For the optical range, NIMs with a negative refractive index were first demonstrated in Ref. 关14兴 where the authors observed the real part of the refractive index n = −0.3 at the telecommunication wavelength of 1.5 m. In that report the authors experimentally verified their early theoretical prediction for negative refraction in an array of parallel metal nanorods 关15兴. The first experimental observation of negative n in the optical range was followed by another successful experiment 关16兴. Note that the losses become progressively important with increasing frequency toward the optic band. Moreover, the elementary cell of the resulting structure is on order of the wavelength. Making a true NIM requires the cell 036609-1 ©2006 The American Physical Society PHYSICAL REVIEW E 73, 036609 共2006兲 SARYCHEV, SHVETS, AND SHALAEV size to be less than / 2. Therefore, miniaturization of the cell size is of major interest, and can be accomplished, for example, by utilizing plasmonic effects 关17–20兴. It is also necessary to take into account from the beginning that dielectric permittivity = ⬘ + i⬙ and magnetic permeability = ⬘ + i⬙ are complex values. The structure that exhibits the negative real dielectric permittivity ⬘ ⬇ −0.7 and negative real of magnetic permeability ⬘ ⬇ −0.3 in green light 共 ⬇ 0.5 m兲 has been investigated in Ref. 关21兴. However, rather large imaginary components 共⬙ ⬇ 1.0 at the resonance兲 has not allowed the observation of the negative refraction. Thus the demonstration of a negative index meta-material in the regime where plasmonic effects are important remains elusive. Plasmonic effects must be correctly accounted for to design a metamaterial with optical magnetism. Below we show that specially arranged metal nanoparticles can support, along with the electrical SPR, magnetic plasmon resonance 共MPR兲. The MPR’s resonance frequency r can be made independent of the absolute characteristic structure size a and / a ⬅ 2c / a. The only defining parameters are the plasmonic permittivity m共兲 and the structure geometry. Such structures act as optical nanoantennas by concentrating large electric and magnetic energies on the nanoscale at the optical frequencies. The magnetic response is characterized by the magnetic polarizability ␣ M with the resonant behavior similar to the electric SPR polarizability ␣E: real part of ␣ M changes the sign near the resonance and becomes negative for ⬎ r, as required for negative index meta-materials. We show that the electrostatic resonances must replace 共or strongly modify兲 GLC resonances in the optical/mid-IR range if a strong magnetic response is desired. The idea of using electrostatic resonances for inducing optical magnetism is relatively recent. For example, electrostatic resonances of periodic plasmonic nanostructures have been recently employed to induce magnetic properties due to close proximity of adjacent nanowires 关17,18兴. Higher multipole electrostatic resonances were shown 关19兴 to hybridize in such a way as to induce magnetic moments in individual nanowires. Strong electrostatic resonances of regularly shaped nanoparticles 共including nanospheres and nanowires兲 occur for −2 ⬍ m ⬘ ⬍ −1. The resistive damping characterized by the ratio m ⬙ / m⬘ of the imaginary and real parts of the dielectric permittivity of a metal is fairly strong for those frequencies corresponding to 兩m ⬘ 兩 ⬃ 1. However, m⬙ / m⬘ is known to decrease for 兩m 兩 Ⰷ 1. Therefore, there is a considerable incentive to design nanostructures exhibiting resonances for m ⬘ Ⰶ −1. Such horseshoe-shaped structures, first suggested in Ref. 关22兴 are described below. Spectrally, these structures support strong magnetic moments at the frequencies higher than the microwave–mid-IR frequencies supported by the traditional split ring resonators 共see Refs. 关3,12兴 for details兲 and lower than the ultraviolet frequencies supported by the sub-wavelength plasmonic crystals described in Refs. 关18,19兴. Conceptually, the horseshoe-shaped structures described here are distinct from the earlier low frequency structures because they are not relying on the GLC resonance for producing a strong magnetic response because plasmonic properties of the metal are very important when the sizes are small and the operational frequencies are high. FIG. 1. Currents in two parallel metal wires excited by external magnetic field H. Displacement currents, “closing” the circuit, are shown by dashed lines. Below we present a three-dimensional theory of the magnetic resonance in a plasmonic structure and related twodimensional numerical simulations. Specifically, here we develop a comprehensive theory and perform detailed numerical simulations for negative index metamaterials based on horseshoe-shaped structures. The possibility of optical magnetism in such structures was first theoretically predicted 关22兴 and recently experimentally verified 关23兴. Closely related split-ring resonator structures were also shown 关24兴 to possess optical magnetism. Here we demonstrate that such horseshoe-shaped structures may have negative dielectric permittivity, in parallel with negative magnetic permeability, and thus they can be used for building a metamaterial with a negative-refractive index. II. ANALYTICAL THEORY OF MAGNETIC PLASMON RESONANCES We first consider a pair of parallel metallic rods. The external magnetic field excites the electric current in the pair of the rods as shown in Fig. 1. The magnetic moment associated with the circular current flowing in the rods results in the magnetic response of the system. Suppose that an external magnetic field H = 兵0 , H0 exp共−it兲 , 0其 is applied perpendicular to the plane of the pair. 共We suppose that magnetic field is along y axis and the rods are in 兵x , z其 plane.兲 The circular current I共z兲 excited by the time-varying magnetic field flows in the opposite directions in the nanowires as shown in Fig. 1. The displacement currents flowing between the nanowires close the circuit. We introduce the “potential drop” U共z兲 = 兰ba Edl between the pair where the integration is along the line 兵a共z兲 , b共z兲其. To find the current I共z兲, we integrate the Faraday’s Law curl E = ik共H0 + Hin兲 over the contour 兵a , b , c , d其 in Fig. 1, where k = / c is a wave vector, and Hin = curlA is the magnetic field induced by the current. It is assumed that the nanowire length 2a is much larger than the distance d between the nanowires and the radius b of a nanowire. We also assume that kd Ⰶ 1. Under these assumptions, the vector potential A is primarily directed along the nanowires 共z direction兲. The use of the integral form of Faraday’s law yields 036609-2 共IR − ik2Az + dU/dz兲⌬z = − ikH0d⌬z, 共1兲 PHYSICAL REVIEW E 73, 036609 共2006兲 MAGNETIC PLASMON RESONANCE where the pair impedance R ⯝ 2 / 共b2兲 ⬇ 8i / 共b2兲, where = 1 + i4 / Ⰶ −1 is the metal complex permittivity, and ±IR / 2 are the electric fields on the surface of the nanowires. Note that the wire resistivity is explicitly taken into account. This sets our calculation apart from the earlier work on the resonances of conducting split ring resonators 关25兴 and conducting stick composites 关2兴 because plasmonic resonances of the wire are now fully accounted for. Electric field E can be always presented in terms of the vector potential A and electric potential as E = − + ikA. In the standard Lorentz gage the electric potential equals to 共r1兲 = 兰exp共ikr12兲q共r2兲 / r12 dr2 and the vector potential A共r1兲 = c−1 兰 exp共ikr12兲j共r2兲 / r12 dr2, where r12 = 兩r1 − r2兩, and q and j are charge and current density correspondingly. In the case of two long wires the currents flow inside the wires. Correspondingly the vector potential A has the only component in the direction z of the wires A = 兵0 , 0 , Az其. Since the vector potential A is perpendicular to the line 兵a共z兲 , b共z兲其 the potential drop U in Eq. 共1兲 equals U共z兲 = 兰ba Edl = a − b, where a and b are the electric potential in the points a共z兲 and b共z兲. We consider the excitation of the antisymmetric mode when the currents in the wires are the same in absolute value but are opposite in the direction 共see Fig. 1兲. Correspondingly the electric charge per unite length Q共z兲 = Qa = −Qb. We assume that the diameter b of the wire is much smaller than the distance d between them and the wire length 2a Ⰷ d. Then the potential drop U共z兲 between the pair estimates as U共z兲 = Q共z兲 / C, where the interwire capacitance C is independent of the coordinate z and is estimated in the Appendix as C ⯝ 关4 ln共d / b兲兴−1. The vector potential Az共z兲 is proportional to the electric current Az共z兲 = 共L / c兲 I共z兲 / 2, where the wire pair inductance is estimated as L ⯝ 4 ln共d / b兲 共see the Appendix兲. Note that both C and L are purely geometric factors that do not depend on the plasmonic nature of the rods. The product LC can be estimated as LC ⯝ 1. Yet, in the above consideration we never assumed that the wires are made from the perfect metal or from a metal with real conductivity, i.e., imaginary permittivity. Moreover, the two wire nanoantenna has most interesting behavior when metal dielectric constant is real and negative. The plasmonic nature of the metal is accounted for below. We substitute U共z兲 and Az共z兲 into Eq. 共1兲, taking into account the charge conservation law dI / dz = iQ共z兲, and obtain a differential equation for the current d2I共z兲 Cd2 2 = − g I共z兲 − H0 , dz2 c 共2兲 where −a ⬍ z ⬍ a , I共−a兲 = I共a兲 = 0, and the parameter g is given as g2 = k2关LC − 8C关共kb兲2m兴−1兴. 共3兲 The two-wire antenna is resonantly excited when G = ga = N / 2, where N is an integer. Note that the material properties of the metal enter the resonant parameter G through the dielectric permittivity m. In the context of the wire pair, the earlier discussed GLC resonances 关3,12,15,25,27兴 correspond to the wire thickness b much larger than the skin depth 共k2 兩 m 兩 兲−1/2. This approximation, typically valid for microwave and mid-IR frequencies, yields g = k / 冑LC and the resonant condition ka = / 2, also known as the antenna resonance. Let’s consider the opposite 共“electrostatic,” as explained below兲 limit of 兩8C关共kb兲2m兴−1 兩 Ⰷ 1. In the electrostatic regime G depends only on the metal permittivity and the aspect ratio G2 ⯝ − 2共a/b兲2 ln共d/b兲/m , 共4兲 but not on the wavelength and absolute length of the wires. Sharp resonance in Eq. 共2兲 requires that G2 be positive, possibly with a very small imaginary part. Indeed, for IR/visible frequencies m is negative 共with a smaller imaginary part兲 for typical 共Ag, Au, etc.兲 low loss metals. Metal dielectric constant m can be approximated by the Drude formula m共兲 ⬵ b − 共 p/兲2/共1 − i/兲, 共5兲 where b is a “polarization” constant, p is the plasma frequency, and = 1 / is the relaxation rate. For considered here silver nanoantennas the constant b approximates as b ⬇ 5, the plasma frequency p ⬇ 9.1 and the relaxation rate ⬇ 0.02 关29兴. For example, at = 1.5 m the silver dielectric constant estimates as m ⬘ ⬇ −120 and m⬙ / 兩m 兩 ⬇ 0.025. We consider now the electric field in the system of two conducting rods still assuming the electrostatic limit when the propagation constant G is given by Eq. 共4兲. The electric charge Q共z兲 and the current I共z兲 关Q共z兲 = 共i兲−1dI共z兲 / dz兴 are given by solution of Eq. 共2兲 Q共z兲 = Q0 I共z兲 = i sin共Gz/a兲 , cos G 冉 共6兲 冊 cos共Gz/a兲 Q 0a 1− , G cos G 共7兲 where Q0 = ibdkH0冑−m / 关4冑2 ln3/2共d / b兲兴. Using the Lorentz gauge we can write the equation for the electric potential 共r兲 = 冕 q共r1兲 exp共ikR1兲 dr1 − R1 冕 q共r2兲 exp共ikR2兲 dr2 , R2 共8兲 where q共r1兲 and q共r2兲 are electric charges distributed over the surface of the rods 1 and 2, R1 = 兩r − r1兩, R2 = 兩r − r2兩, and the integration goes over the rods 1 and 2. We consider the electric field between the rods, i.e., in 兵z , x其 plane 共see Fig. 1兲 and assume that 兩x 兩 Ⰶ a, 兩z 兩 ⬍ a, and the distances to the rods d1 = 兩x − d / 2 兩 Ⰷ b and d2 = 兩x + d / 2 兩 Ⰷ b. Then we can integrate in Eq. 共8兲 over the cross section of the rod after which it takes the one-dimension form 共x,z兲 = 冕 a −a Q共z1兲 冋 册 exp共ikR1兲 exp共ikR2兲 − dz1 , R1 R2 共9兲 where the linear charge density Q共z1兲, obtained from q共r1兲 by 036609-3 PHYSICAL REVIEW E 73, 036609 共2006兲 SARYCHEV, SHVETS, AND SHALAEV integration over the rod circumference, is given by Eq. 共6兲 关Q共z1兲 = −Q共z2兲兴, the distances R1 and R2 take the form R1 = 冑d21 + 共z − z1兲2, R2 = 冑d22 + 共z − z1兲2. Two terms in the square brackets in Eq. 共9兲 cancel when 兩z − z1 兩 ⬎ d, as it is discussed in the Appendix. Since we assume that kd Ⰶ 1 and d Ⰶ a we can put the exponents exp共ikR1兲 ⯝ exp共ikR1兲 ⯝ 1 and extend the integration in Eq. 共9兲 from z1 = −⬁ to z1 = ⬁. Resulting integral is solved explicitly and we obtain the analytical equation for the electric potential in the system of two nanowires 冏 共x,z兲 = 2 ln 冏 d/2 + x Q共z兲, d/2 − x 共10兲 where Q共z兲 is given by Eq. 共6兲. Extrapolation of this result to the surface of the wires gives the potential drop U共z兲 = 共d / 2 − b , z兲 − 共−d / 2 + b , z兲 = 4 ln共2d / b兲Q共z兲. Thus we obtain that the interwire capacitance C = Q共z兲 / U共z兲 =4 ln共d / b兲 is a constant, which is independent of the coordinate z, in agreement with estimate done in the Appendix. The vector potential A = 兵0 , 0 , A其 is calculated in a similar way A共x,z兲 ⯝ ⯝ 1 c 1 c 冕 冋 冕 冋 册 I共z1兲 exp共ikR1兲 exp共ikR2兲 − dz1 R1 R2 I共z1兲 1 2 1 d/2 + x − dz1 ⯝ ln I共z兲, R1 R2 c d/2 − x a −a ⬁ −⬁ 册 冏 冏 共11兲 where the electric current is given by Eq. 共7兲. Extrapolating to the vector potential A to the surface of the first wire 共x = d / 2 − b兲 we obtain 2cA = LI, where the interwire inductance L equals to L ⯝ 4 ln 共d / b兲. The inductance L is also independent of the coordinate z in agreement with the Appendix. Since the interwire capacitance C and inductance L both remain constant along the wires, the Maxwell equations reduce to an ordinary differential Eq. 共2兲. The electric field E = − + ikA is calculated from the potentials 共10兲 and 共11兲 as Ex = − Ez = − 冏 2Q0 d/2 + x ln aG d/2 − x 再 冉冊 ⫻ G2 cos 冉 冊 2Q0d Gz sin sec共G兲, 共d/2兲2 − x2 a 冏 共12兲 共13兲 冋 冉 冊 册冎 Gz Gz sec共G兲 − a2k2 1 − cos sec共G兲 a a , 共14兲 where we still assume that 兩x 兩 Ⰶ a, 兩z 兩 ⬍ a, 兩x − d / 2 兩 Ⰷ b, and 兩x + d / 2 兩 Ⰷ b. The transverse electric field Ex changes its sign with the coordinate z vanishing at z = 0. Yet, on average the ratio 兩Ex 兩 / 兩Ez兩 is estimated near the resonance 共G ⬇ / 2兲 as 兩Ex 兩 / 兩Ez 兩 ⬃ a / d Ⰷ 1, that is the transversal electric field is on average much larger than the longitudinal field at MPR. Near the wires transverse field Ex increases even more: 兩Ex 兩 ⬃Q0 / b. The potential drop ⌬U between the points x1 = d / 2 + l / 2 and x2 = d / 2 − l / 2 共2b ⬍ l Ⰶ d兲 is estimated from Eq. 共13兲 as −2lQ共z兲 / d, and the corresponding electric field Eout ⯝ − 2Q共z兲 / d. This field should be considered an external field for the wire at the coordinate y = d / 2. The internal transverse potential drop across the wire is estimated as 兩Uin 兩 ⯝ b 兩 Eout 兩 / 兩m 兩 ⯝ 2 兩 Q共z兲 兩 b / 共d 兩 m 兩 兲, where m is the metal dielectric constant, which is assumed to be large 兩m 兩 Ⰷ 1. The problem of the internal transverse field closely resembles the classical problem of the field induced in a dielectric cylinder by another charged cylinder placed parallel to the first cylinder. An elegant solution of the problem can be found in Ref. 关26兴, Sec. 7. In the discussed case 兩m兩, d / b Ⰷ 1 it gives the above obtained estimate for Uin.The ratio of the potential drop Uin across a wire to the potential drop U共z兲 between the wires equals 冏 冏 Uin共z兲 b Ⰶ 1. ⯝ U共z兲 d兩m兩ln共d/b兲 共15兲 For any practical purpose we can neglect Uin in comparison to U, which allows us to reduce the problem of charge and current distribution in the two wire system to the solution of the ordinary differential equation 共2兲 for the electric current I共z兲. Condition 共15兲 is important for the developed analytical theory of MPR in the system of two thin rods. However, we can envision a system 共e.g., two closely packed metal nanowires or hemispheres兲 that still reveals MPR, but condition 共15兲 is not applied. To clarify the nature of the resonance, it is instructive to compute the ratio of the electric and magnetic energies at the resonance EE EM 冕 冕 C−1 ⬃ c2 L 兩Q共z兲兩2dz ⬇ 兩I共z兲兩2dz 2 g2 , 2 ⬇1− k ln共d/b兲k2b2兩m兩 共16兲 where we assume that the spatial frequency g, given by Eq. 共3兲, is close to the resonance 共ga ⬇ / 2兲 and use the expressions for the specific capacitance C ⯝ 关4 ln共d / b兲兴−1 and inductance L ⯝ 4 ln共d / b兲 derived in the Appendix. In the electrostatic limit 兩8C关共kb兲2m兴−1 兩 Ⰷ 1 we obtain UE / U M Ⰷ 1 thus explaining the name given to this regime. Because of the symmetry of the electric potential considered here, it is clear that such polarization cannot be induced by any uniform electric field. Therefore, the discussed resonance can be classified as the dark mode 关31兴. The electric current I共z兲 is found from Eq. 共2兲 and used to calculate the magnetic moment of the wire pair m = 共2c兲−1 兰 关r ⫻ j共r兲兴dr, where j共r兲 is the density of the current and the integration is over the two nanowires. Thus we obtain tan G − G 1 . m = H0a3 ln共d/b兲共kd兲2 2 G3 共17兲 The metal permittivity m has a large negative value in the optical–near-IR range while its imaginary part is small; 036609-4 PHYSICAL REVIEW E 73, 036609 共2006兲 MAGNETIC PLASMON RESONANCE therefore, the magnetic moment m has a resonance at G ⬇ / 2 关see Eq. 共3兲兴 when m attains large values. For a typical metal we can use the Drude formula 共5兲 for m, where the relaxation parameter is small / Ⰶ 1. Then the normalized magnetic polarizability ␣ M has the following form near the MPR: ␣M = 16ad p 4m 关1 − /r − i /共2r兲兴−1 , = 2 H0V r冑2 ln共d/b兲 共18兲 r where V = 4abd, and MPR frequency = b p冑2 ln共d / b兲 / 共4a兲. Note that the magnetic moment contains a prefactor 2da / c2 Ⰶ 1 that is small in the electrostatic limit valid for subwavelength nanostructures, as was predicted earlier 关19兴. Close to the resonance 共G = / 2兲 the enhancement factor can be very large for optical and infrared frequencies because of the high quality of the plasmon resonance for r Ⰶ p. Therefore, the total prefactor in Eq. 共18兲 can be of the order of one, thereby enabling the excitation of a strong MPR. Although the electric field energy near resonance is also very high, it is primarily concentrated in the perpendicular to the wires component of the electric field connecting the two wires as it was discussed after Eqs. 共12兲 and 共13兲. If the wave vector of the propagating wave is in the plane of the wires, and perpendicular to the wires, then the described above MPR does not strongly affect the electric field component parallel to the wires. The integral from the electric field, which is generated in magnetic resonance between the wires, exactly equals to zero as it follow from Eqs. 共12兲 and 共13兲. Envisioning a composite material that consists of such wire pairs, we expect that the magnetic plasmon resonance will not contribute to the dielectric permittivity in the direction parallel to the wires. Therefore, such a medium is not bianisotropic 关25兴 and can be described by two effective separate parameters and . Prior work on inducing magnetic moments in nanostructured materials 关18,19兴 in the electrostatic regime dealt with highly symmetric 共round兲 nanowires. Here we demonstrate that magnetic moment can also be induced in the strongly nonsymmetric 共specifically, horseshoeshaped兲 nanostructures with a large negative value of . We now consider a metal nanoantenna that has a horseshoe shape, which is obtained from a pair of nanowires by shorting it at one of the ends 共see Fig. 2兲. When the quasistatic condition 兩8C关共kb兲2m兴−1 兩 Ⰷ 1 holds, the electric current I共z兲 in a horseshoe nanoantenna can be obtained from Eq. 共2兲, where the boundary condition changes to Iz=a = 共dI / dz兲z=0 = 0 and, as above, a Ⰷ d Ⰷ b. It is easy to check that the magnetic polarizability ␣ M is still given by Eq. 共18兲, where a is now equal to the total length of the horseshoe nanoantenna. Therefore, the horseshoe nanoantenna provides the same magnetic polarizability ␣ M at the twice shorter length. Magnetic permeability = 1 + i2 for a metamaterial where the silver horseshoe nanoantennas are oriented in one direction 共“z” direction in Fig. 1兲 and are organized in the periodic square lattice is shown in Fig. 2; the optical parameters for silver were taken from Ref. 关29兴. As one can see in the figure, the negative magnetism can be observed, for ex- FIG. 2. Optical magnetic permeability = 1 + i2 共1: continous line, 2: dashed line兲 estimated from Lorenz-Lorentz formula for the composite containing shaped silver nanoantennas; volume concentration p = 0.3; left curves: a = 200 nm, d = 50 nm, b = 13 nm; right curves a = 600 nm, d = 90 nm, b = 13 nm. ample, in the near-infrared part of the spectrum, including the telecommunication wavelength of 1.5 m. III. NUMERICAL SIMULATIONS OF TWO-DIMENSIONAL STRUCTURES To obtain a magnetically active in the optical range material, it might be more convenient to employ 共and much easier to model兲 a “two-dimensional” metamaterial, with the nanoantennas having a horseshoe shape in x , y plane and infinitely extended in the z direction. When the quasistatic condition 兩k2bdm 兩 Ⰶ 1 holds 共b and d are thickness and distance between the opposite walls, correspondingly兲, the MPR frequency r is defined by the equation G2 = 2a冑−2 / 共mbd兲 = / 2. The resonant magnetic field is shown in Fig. 3. The finite elements code FEMLAB 关28兴 was used to calculate the field distribution. Note that the magnetic field inside the horseshoe is large and of the opposite sign with the external field H0, resulting in a negative magnetic permeability in a close proximity to the magnetic plasmon resonance. To estimate the effective magnetic permeability we use an earlier developed approach 关13兴, which gives z = 1 + p共sH0兲−1 兰 共Hin − H0兲ds = 共32/ 兲a2 p−2共 / 2 − G2兲−1 for a plasmonic crystal composed by the horseshoe nano antennas, where Hin is the magnetic field inside a horseshoe, the inte- FIG. 3. 共Color online兲 Magnetic plasmon resonance in a silver horseshoe-shaped nanoantenna placed in a maximum of external field H0 which is directed perpendicular to the plane; the frequency corresponds to = 1.5 m; silver dielectric constant is calculated from the Drude formula 共5兲. 036609-5 PHYSICAL REVIEW E 73, 036609 共2006兲 SARYCHEV, SHVETS, AND SHALAEV FIG. 4. Optical magnetic permeability = 1 + i2 共1: continuous line, 2: dashed line兲 of the composite containing silver nanoantennas shown in Fig. 3 organized in a square lattice; volume concentration p = 0.4. gration is over the area s = da, p is the concentration of the nanoantennas organized in a square lattice. For a good optical metal, such as gold or silver, z becomes large and negative for ⬎ r. For example, the magnetic permeability 共both real and imaginary parts兲 of a metamaterial composed of silver horseshoes has a sharp resonance for = r shown in Fig. 4. The real part 1 of the optical magnetic permeability turns negative for ⬎ r. Figure 4 also reveals the spectral range 艌 1 m where 1 is still less than −1 while the relative losses are small: ␦ = 2 / 1 Ⰶ 1. The losses are crucial for a such application of NIM as the perfect lens. Losses could, in principle, be further reduced by cooling the metal nanoantennas to cryogenic temperatures. Simple estimates show that even at the liquid nitrogen temperature the electron mean free path becomes of the order of the horse shoe size. The optical properties of metals are not well understood when the mean free path becomes larger than the nanoantenna size. For these reasons we only consider metamaterials at room temperature. To illustrate the effective magnetic properties of the horseshoe metamaterials, we simulated em wave propagation in the plasmonic crystal composed from silver nanoantennas shown in Fig. 3. 关The first three horseshoe columns are well seen in Fig. 5共a兲.兴 The em wave is incident on the crystal from the left, its vacuum wavelength is = 1.4 m. The wave is evanescent in the crystal, as can be clearly seen from Fig. 5, and the transmittance T ⬍ 10−6. This evanescence is due to the negative magnetic permeability of the crystal that could be rather large at the resonance 共see Fig. 4兲. One way of making this crystal transparent is to fill the space between the columns of the horseshoes by a material with negative dielectric constant = −3. We speculate that such a modification can lead to a double-negative metamaterial. This is confirmed by the numerical simulations. The results are shown in Fig. 5 共right兲. Indeed, the addition of a negative- material turns our negative- metamaterial into transparent NIM. Note that the negative- material was added only between the adjacent columns of horseshoe-shaped nanoantennas. No additional material was placed in the exterior of the nanoantenna. This was done intentionally because modification of the nanostructure region where most of the magnetic field is concentrated is known 关30兴 to affect the permeability effective permittivity. Interestingly, the horseshoes themselves can exhibit a double-negative behavior when they are closely packed. We designed a two-dimensional dense periodic structure consist- FIG. 5. 共Color online兲 Wave propagation through a twodimensional 共infinitely extended in the normal to the page direction兲 plasmonic crystal near the plasmon magnetic resonance. The crystal is composed of the u-shaped nanoantennas shown in Fig. 3, volume filling ratio p = 0.4, = 1.4 m. Magnetic field H of the incident em wave is directed along z axis perpendicular to plane of picture; Hz = 1 in the incident wave. 共Top兲 Spaces between the horseshoes are filled with vacuum: no propagation. 共Bottom兲 Spaces between the horseshoes are filled with a hypothetical = −3 material: wave propagates into the structure. ing of alternating up and down horseshoe nanoantennas. One half of the elementary cell is shown in Fig. 6共a兲; another half of the elementary sell is obtained by 180ⴰ degree rotation in the xy plane. 共The structure then repeats itself in x and y directions; separation between antenna centers is 80 nm, horizontal periodicity is 160 nm, vertical periodicity is 400 nm; see Fig. 5.兲 The dispersion relation 共kx兲 for the electromagnetic wave propagating through the periodic structure in x direction has been calculated by numerically solving Maxwell’s equation for magnetic field Hz. For computational simplicity, we have assumed a hypothetical lossless plasmonic material with the frequency-dependent dielectric permittivity = 1 − 2p / 2, where 2c / p = 225 nm. The frequency and the wave vector k are normalized to 0 = 2c / 0 and k0 = 2 / 0, respectively, where 0 = 1.5 m. Remarkably, one of the propagating modes 关shown in Fig. 6共b兲兴 exhibits the left-handedness: its group velocity vgr = / k opposes its phase velocity. Figure 6共a兲 shows the magnetic field profile and the electric field inside the elementary cell for kx = 0 共magnetic cutoff condition corresponding to = 0兲. Magnetic field is concentrated inside the horseshoes, and has opposite signs in the adjacent horseshoes. The dominant field in the structure is Ex which does not contribute to the Poynting flux in the propagation direction. Note that arrows in Fig. 6 indicate the value the electric field takes at their origin. Then Fig. 6 clearly indicates that the electric 036609-6 PHYSICAL REVIEW E 73, 036609 共2006兲 MAGNETIC PLASMON RESONANCE FIG. 6. 共Color online兲 Plasmonic crystal composed from the u-shaped metal nanoantennas; separation between antenna centers is 80 nm. Magnetic 共color and contours兲 and electric 共arrows兲 fields inside a periodic array of horseshoe-shaped nanoantennas at the cutoff 共kx = 0兲. 共b兲 Dispersion relation vs kx for a left-handed electromagnetic wave. field of MPR is mainly confined inside the horseshoes and is almost negligible in the metal. As we have mentioned above there is half of the elementary cell in Fig. 6. In another half cell the Ex field has opposite direction so that the average Ex is zero. The electric field is primarily potential 共i.e., can be derived from an electrostatic potential兲, but has a nonvanishing solenoidal component that produces the magnetic field. The potential drop between the metal arms is much larger than the potential drop between external and internal interfaces of an arm. This behavior of the MPR electric field, obtained from direct computer simulations, resembles our analytical results for two wire system 关see discussion at Eqs. 共12兲 and 共13兲兴. The fact that the dominant electric field Ex does not change the sign inside the cell along the direction of the propagation indicates that the mode in question does not owe its negative dispersion to the band-folding effect common in photonic crystals. The left-handed behavior occurs in the vicinity of = 1.88 m, which is close to the MPR resonance. The negative necessary for the negative refractive index is induced by the proximity of the dipole-type electrostatic resonances 关18兴. nances in the dielectric permittivity and magnetic permeability . ACKNOWLEDGMENTS The authors acknowledge useful contributions and discussions with D. Genov, V. Podolskiy, and G. Tartakovsky. This work was supported in part by NSF Grants Nos. ECS0210445 and DMR-0121814 and by the ARO MURI Grant No. W911NF-04-01-0203. APPENDIX We derive equations for the capacitance C and inductance L between two parallel wires of the radius b and length a separated by the distance d. We suppose from the beginning that b Ⰶ d and d Ⰶ a. We also suppose that the dielectric constant m of the wires is large in absolute value 兩m 兩 Ⰷ 1 whereas the skin depth ␦ ⬃ 共k冑m兲−1Ⰷb as it is explained in the text. To find the capacitance C we first calculate the electric potential ⌽a in the point with coordinate ra at the surface of the wire 共point a in Fig. 1兲 obtaining ⌽a = IV. CONCLUSIONS In conclusion, the phenomenon of a magnetic plasmon resonance in metallic horseshoe-shaped split rings was described. This resonance is distinctly different from the geometric LC resonance described earlier for split rings because it is determined by the plasmonic properties of the metal. This work paves the way to designing metallic metamaterials that are magnetically active in the optical and near-infrared spectral ranges. Presented three-dimensional analytic calculations and two-dimensional numerical simulations reveal that resonantly enhanced magnetic moments can be induced in very thin 共thinner than a skin depth兲 split rings with typical dimensions much shorter than the wavelength 共on the order of 100 nanometers兲. Periodic arrays of such horseshoeshaped nanoantennas can be used to design left-handed metamaterials by exploiting the proximity of electric reso- 冕 q1共r1兲 exp共ikra1兲 dr1 + ra1 冕 q2共r2兲 exp共ikra2兲 dr2 , ra2 共A1兲 where ra1 = 兩ra − r1兩, ra2 = 兩ra − r2兩, q1 and q2 are the electric charges distributed over the surface of the rods; the integration goes over the surface of the first 共a , d兲 and second 共b , c兲 rods in Fig. 1. For further consideration we choose the coordinate system 兵x , y , z其 with z axis along the 共a , d兲 rod, origin in the center of the system and the x axis connecting the axes of the rods so that y axis is perpendicular to the plane of two rods. We introduce the vector d = 兵d , 0 , 0其 between the wires and two-dimensional unit vector 共兲 = 兵cos , sin 其 in 兵x , y其 plane, where is the polar angle. Then the vectors in Eq. 共A1兲 can be written as ra共a , za兲 = 兵b共a兲 + d / 2 , za其, and r2共2 , z2兲 = 兵b共2兲 r1共1 , z1兲 = 兵b共1兲 + d / 2 , z1其, − d / 2 , z2其. It follows from the symmetry of the problem that 036609-7 PHYSICAL REVIEW E 73, 036609 共2006兲 SARYCHEV, SHVETS, AND SHALAEV the electric charge q1共 , z兲 = −q2共 + , z兲 共recall that we consider antisymmetric mode when the electric currents in the rods are equal in absolute values but follows in the opposite directions兲. We rewrite Eq. 共A1兲 splitting it in two parts 共1兲 ⌽a ⬅ ⌽共0兲 a + ⌽a , ⌽共0兲 a 共za, a兲 = − ⌽共1兲 a 冕 冕 =0 冑⌬z ⬅ a q共,z兲 + 共b⌬2 − d兲2 冕 冕 冋冑 1 2 共A3兲 dzbd , 共A4兲 z=−a 1 2 共A2兲 ⌬z + b2⌬21 q共r2兲 exp共ikra2兲 − 1 dr2 , ra2 共A5兲 where ⌬z = za − z, ⌬1 = a − = 兵cos a − cos , sin a − sin 其, ⌬2 = a + = 兵cos a + cos , sin a + sin 其. Note that dimensionless vectors ⌬1 and ⌬2 satisfy 兩⌬1兩, 兩⌬2 兩 ⬍ 2 so that the second terms in the radicals in Eq. 共A3兲 are much less than a. The electric currents in the rods and, correspondingly, electric charge q changes with coordinate z on the scale ⬃a which is much larger than the distance d between the rods. Therefore we can neglect z variation of the electric charge for 兩⌬z 兩 ⬍ d. On the other hand the term in the square brackets in Eq. 共A3兲 vanishes as ⬃d2 / 兩⌬z兩3 for 兩⌬z 兩 ⬎ d. This allows to replace in Eq. 共A3兲 the charge q共z , 兲 by its value q共za,a兲 in the observation point ra obtaining ⌽共0兲 a 共za, a兲 = − 冕 2 =0 q共za,兲 冕 a z=−a 冋冑 1 ⌬z + b2⌬21 2 共A6兲 1 冑⌬z2 + 共b⌬2 − d兲2 dzbd; 冉 2 冊 共A9兲 The second term ⌽共1兲 a in Eq. 共A2兲 is small in the limit of a Ⰶ , i.e., kra1, kra2 Ⰶ 1. The real part of ⌽共1兲 a gives a small correction ⬃共d / a兲2 to the potential ⌽共0兲 that can be neglect. a The imaginary part is important regardless of its absolute value since it gives so-called radiative losses. To estimate the losses we assume that b / d Ⰶ 1 and neglect the angle dependence of the charge distribution. Then we expand Eq. 共A5兲 in series of k and linearly approximate Q共z兲 ⯝ q1z 关recall that Q共z兲 is an odd function of z兴 obtaining 共A7兲 d = Q共z兲arcosh − 1 + O„共d/a兲2…, 2b2 2 1 ⯝ 2 arcosh关共d/b兲2/2 − 1兴 − i 共ak兲3共kd兲2 45 C 冉冊 2 d − i 共ak兲3共kd兲2 , 45 b ⯝4 ln 共A11兲 共A12兲 where the first term is the capacitance between two parallel infinite cylinders 共see Ref. 关26兴, Chap. 3兲; the second term gives the radiative losses due to the retardation effects. Consider now the inductance L between the wires. To find the inductance L we first calculate the vector potential Aa in the point with coordinate ra inside the wire. We neglect the edge effects and assume that the vector potential is parallel to the axes of the wires obtaining 1 c 冕 冉 j共r兲 冊 exp共ikra1兲 exp共ikra2兲 − dr, ra1 ra2 共A13兲 where ra1 = 兩ra − r兩 and ra2 = 兩ra − r + d兩, j共r1兲 is the density of the current and the integration goes over the volume of the first wire. We consider the quasistatic case when the skin effect is small 共kb冑兩m兩 Ⰶ 1兲. Then the electric current uniformly distributes over the cross section of a wire and j共r兲 = I共z兲 / 共b2兲. Following the procedure used above for calculating the electric potential, the vector potential is expressed 共1兲 as Aa = A共0兲 a + Aa , where A共0兲 a = A共1兲 a = 共A8兲 where the second term includes all corrections to the integral 共A3兲 due to finite size of the system. For the thin wires, considered here, when the radius b is much smaller than the 共A10兲 where we neglect the terms with higher orders on k as well as all terms on the order of 共bk兲2. Due to the symmetry of the system the potential difference U = ⌽a − ⌽b between points a and b 共see Fig. 1; za = zb兲 equals to U = 2⌽a. The capacitance C defined as C = Q共z兲 / U共z兲 is given by Aa = the accuracy of this replacement is about 共d / a兲2 Ⰶ 1. Since we consider the quasistatic limit when the distance between the rods d Ⰶ and the metal dielectric constant 兩m 兩 Ⰷ 1 the potential lines in 兵x , y其 plane are close to the static case. Therefore we can safety suppose that the angle distribution of the electric charge q共z , 兲 is the same as it would be in the case of two infinite metal cylinders in the static case: q共z , 兲 = Q共z兲冑共d / b兲2 − 4 / 关2共d + 2b cos 兲兴, where Q共z兲 is the electric charge per unit length of the rod so that =2 q共z,兲bd = Q共z兲. Then the integral in Eq. 共A6兲 gives 兰=0 the ⌽共0兲 a 共z兲 ⌽共0兲 a 共z兲 ⯝ Q共z兲2关ln共d/b兲兴. 3 2 ⌽共1兲 a 共z兲 ⯝ − iQ共z兲共ak兲 共kd兲 /45, exp共ikra1兲 − 1 q共r1兲 dr1 ra1 − distance d between the wires the potential ⌽共0兲 a approximates as 冕 冉 1 c 冕 冉 冊 1 I共z兲 1 − dr, 2 b ra1 ra2 共A14兲 冊 I共z兲 exp共ikra1兲 − 1 exp共ikra2兲 − 1 − dr. b2 ra1 ra2 共A15兲 共0兲 The term A共0兲 a estimates in the same way as ⌽a . As result we 共0兲 obtain the vector potential Aa averaged over the crosssection of the wire in the following form: 036609-8 PHYSICAL REVIEW E 73, 036609 共2006兲 MAGNETIC PLASMON RESONANCE A共0兲 a 共z兲 ⯝ 冋 冉冊 册 d I共z兲 4 ln +1 , 2c b 共A16兲 where I共z兲 is electric current, and we neglect terms on the order 共b / d兲2 Ⰶ 1 and 共d / a兲2 Ⰶ 1. To estimate Eq. 共A15兲 we expand it in series on k obtaining that the linear term equals to zero, the k2 term gives small correction 关⬃共kd兲2兴 to A共0兲 a and the third order on k gives the radiative losses, namely, 2 A共1兲 a ⯝ i共kd兲 k 1 c 冕 I共z兲dz ⬃ 2i I共z兲 共kd兲2ka, c 共A17兲 where we rather arbitrary neglect variation of the current over the rod length in transition to the second estimate. We obtain inductance L form the equation Aa − Ab = 2Aa = 共L / c兲I共z兲 as We are now in a position to compare the radiation losses 共given by imaginary parts of capacitance C and inductance L兲 and the ohmic loss in the metal wires. In near infrared spectral region the dielectric constant m for a “good” optical meal 共Ag, Au, etc.兲 can be estimated from the Drude formula 共5兲 as m共兲 ⬃ 共 p / 兲2共1 − i / 兲−1, where p is plasma frequency and Ⰶ Ⰶ p is the relaxation rate. Thus we obtain that the real part of the rod resistance ROhm ⬃ 8共 / 2p兲共a / b2兲 should be compared with “radiation” resistance Rrad ⬃ 共kd兲2共ka兲2 / c. For the silver nanowires, considered in the paper, the Ohmic losses either larger 共ROhm ⬎ Rrad兲 or much larger 共ROhm Ⰷ Rrad兲 than the radiation losses. Therefore we can neglect the imaginary parts ofthe capacitance C and inductance L and approximate them for simplicity as 冉冊 L = 4 ln d + 1 + 4i共kd兲2ka. b L⯝ 1 d ⯝ 4 ln . C b 共A18兲 The first two terms correspond to the self-inductance per unit length of a system of two parallel infinite wires 共Ref. 关26兴, Chap. 34兲. This estimate as well as Eq. 共A12兲 are certainly invalid near the ends of the rods, but in calculating the current distribution I共z兲 and magnetic moment this region is unimportant. This estimate is of logarithmic accuracy; its relative error is on the order of 共4 lnd / b兲−1. Note that the radiation losses crucially depend on the parameter ka. Magnetic plasmon resonance address in this paper becomes very broad when ka ⬎ 1, placing a rather sever constraint on the length 2a of the wire. 关1兴 J. B. Pendry, A. J. Holden, W. J. Stewart, and I. Youngs, Phys. Rev. Lett. 76, 4773 共1996兲. 关2兴 A. N. Lagarkov, A. K. Sarychev, Y. R. Smychkovich, and A. P. Vinogradov, J. Electromagn. Waves Appl. 6, 1159 共1992兲; A. N. Lagarkov and A. K. Sarychev, Phys. Rev. B 53, 6318 共1996兲. 关3兴 J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J. Stewart, IEEE Trans. Microwave Theory Tech. 47, 2075 共1999兲; M. C. K. Wiltshire, J. V. Hajnal, J. B. Pendry, D. J. Edwards, and C. J. Stevens, Opt. Express 11, 709 共2003兲. 关4兴 V. G. Veselago, Sov. Phys. Usp. 10, 509 共1968兲. 关5兴 D. R. Smith, W. J. Padilla, D. C. Vier, S. C. Nemat-Nasser, and S. Schultz, Phys. Rev. Lett. 84, 4184 共2000兲. 关6兴 J. B. Pendry, Phys. Rev. Lett. 85, 3966 共2000兲. 关7兴 For recent references see J. Pendzy, Opt. Express 11, 639 共2003兲. 关8兴 A. A. Houck, J. B. Brock, and I. L. Chuang, Phys. Rev. Lett. 90, 137401 共2003兲. 关9兴 C. G. Parazzoli, R. B. Greegor, K. Li, B. E. C. Koltenbah, and M. Tanielian, Phys. Rev. Lett. 90, 107401 共2003兲. 关10兴 A. Alu and N. Engheta, IEEE Trans. Microwave Theory Tech. 52, 199 共2004兲. 关11兴 A. N. Lagarkov and V. N. Kissel, Phys. Rev. Lett. 92, 077401 共2004兲. 关12兴 S. Linden, C. Enkrich, M. Wegener, J. Zhou, T. Koschny, and C. M. Soukoulis, Science 306, 1351 共2004兲; N. Katsarakis, T. Koschny, M. Kafesaki, E. N. Economy, and C. M. Soukoulis, Appl. Phys. Lett. 84, 2943 共2004兲. 关13兴 A. K. Sarychev and V. M. Shalaev, Phys. Rep. 333, 275 共2000兲. 关14兴 V. M. Shalaev, W. Cai, U. Chettiar, H.-K. Yuan, A. K. Sarychev, V. P. Drachev, and A. V. Kildishev, physics/0504091, Apr. 13, 共2005兲; Laser Phys. Lett. 3, 49 共2006兲; Opt. Lett. 30, 3356 共2005兲. 关15兴 V. A. Podolskiy, A. K. Sarychev, and V. M. Shalaev, J. Nonlinear Opt. Phys. Mater. 11, 65 共2002兲; Opt. Express 11, 735 共2003兲; A. K. Sarychev, V. P. Drachev, H.-K. Yuan, V. A. Podolskiy, and V. M. Shalaev, Proc. SPIE 5219, 92 共2003兲. 关16兴 S. Zhang, W. Fan, N. C. Panoiu, K. J. Malloy, R. M. Osgood, and S. R. J. Brueck, Phys. Rev. Lett. 95, 137404 共2005兲. 关17兴 G. Shvets, Phys. Rev. B 67, 035109 共2003兲. 关18兴 G. Shvets and Ya. A. Urzhumov, Phys. Rev. Lett. 93, 243902 共2004兲. 关19兴 G. Shvets and Ya. A. Urzhumov, J. Opt. A, Pure Appl. Opt. 7, S23 共2005兲. 关20兴 G. Shvets and Ya. A. Urzhumov, J. Opt. A: Pure Appl. Opt. 共to be published兲. 关21兴 A. N. Grigorenko, A. K. Geim, H. F. Gleeson, Y. Zhang, A. A. Firsov, I. Y. Khrushchev, and J. Petrovic, Nature 共London兲 438, 335 共2005兲. 关22兴 A. K. Sarychev and V. M. Shalaev, Proc. SPIE 5508, 128 共2004兲. 关23兴 C. Enkrich, M. Wegener, S. Linden, S. Burger, L. Zschiedrich, F. Schmidt, J. Zhou, Th. Koschny, and C. M. Soukoulis, condmat/0504774. 关24兴 T. J. Yen, W. J. Padilla, N. Fang, D. C. Vier, D. R. Smith, J. B. Pendry, D. N. Basov, and X. Zhang, Science 203, 1494 共2004兲. 关25兴 R. Marques, F. Medina, and R. Rafii-El-Idrissi, Phys. Rev. B 65, 144440 共2002兲. 036609-9 PHYSICAL REVIEW E 73, 036609 共2006兲 SARYCHEV, SHVETS, AND SHALAEV 关26兴 D. Landau and E. M. Lifshitz, Electrodynamics of Continuous Media, 2nd ed. 共Pergamon, Oxford, 1984兲. 关27兴 L. V. Panina, A. N. Grigorenko, and D. P. Makhnovskiy, Phys. Rev. B 66, 155411 共2002兲. 关28兴 FEMLAB Reference Manual 2003 Version 2.3 Comsol AB, Sweden. 关29兴 U. Kreibig and M. Vollmer, Optical Properties of Metal Clus- ters 共Springer-Verlag, Berlin, 1995兲; P. B. Johnson and R. W. Christy, Phys. Rev. B 6, 4370 共1972兲. 关30兴 A. L. Pokrovsky and A. L. Efros, Phys. Rev. Lett. 89, 093901 共2002兲. 关31兴 M. I. Stockman, S. V. Faleev, and D. J. Bergman, Phys. Rev. Lett. 87, 167401 共2001兲. 036609-10